Maneuvering target tracking using the raw data with unknown threshold under low Signal-To-Noise Ratio (SNR) circumstance is a difficult problem.In this paper, an Interactive Multiple Model (IMM) Bernoulli Track-Before-Detect (TBD) algorithm is proposed.The algorithm uses the IMM method to predict the sampling particles in each target state of the filter, the Bernoulli filter for the recursion of the target particles, and the TBD algorithm in the particle updating phase, thus to realize the updating estimation of the existence probability and distribution density of the targets.In the particle prediction, several models are used to transfer the prediction, which makes the motion state of the predicted particles similar to that of the real targets.The method combines the features of the Bernoulli TBD algorithm with those of the IMM method, which can be used for the high-accuracy detection and tracking of maneuvering targets under the condition of low SNR.Simulation results show that the proposed filter can estimate the target position in real time, and has better filtering performance than the traditional Bernoulli TBD algorithm.